Herding behaviour and price convergence clubs in cryptocurrencies during bull and bear markets

Abstract

This paper sets out to explore whether convergence and herding phenomena exist for digital currencies. Daily data cover a large spectrum of cryptocurrencies in separate bull and bear periods. Empirical estimations for detecting club convergence and clustering are performed by the methodology proposed by Phillips and Shu (2007, 2009). Econometric outcomes reveal preliminary evidence of powerful herding behaviour. The lowest of large-cap digital currencies attract mainly cryptocurrencies which are about in the middle of the large-cap and medium-cap categories whereas the highest-cap cryptocurrencies are parts of convergence clubs with mostly large-cap or purely medium-cap digital currencies during bear markets. Notably, segmentation is higher during bear markets as clusters are formed around more numerous cryptocurrencies than during bull markets. Convergence is stronger during flourishing periods. Secondary herding is also realized among pairs of clubs. Our findings enable investors to better diversify their portfolios and ameliorate their risk-return trade-off during extreme events.

Introduction

Cryptocurrencies have received increasing attention by policymakers, academics, investors, and the media since the astonishingly rapid growth in their returns during 2017. Their innovative features such as decentralization in payment systems, lower transaction costs, pseudonymity, and speed of transactions have increased broad popularity as an economic unit of account (Böhme et al., 2015). Bitcoin has been characterized as a form of liquidity, placed between that of gold and fiat money, and is considered to satisfy the means of transaction function at a competent level. Moreover, it is argued that low transaction fees for digital currency payments are driven to a large extent by a subsidy that is attributed to miners in the form of new currency. Nevertheless, the majority of existing cryptocurrencies display a pre-determined future path supply. Thereby the total coins offered will be fixed. Problems that digital currencies deal with include that of double-spending, which is dealt with by employing peer-to-peer networks and open-source software (Dwyer, 2015).

A significant stream of literature has investigated the effects between financial assets of major importance (Beckmann and Czudaj, 2013, Das et al., 2019, El Abed and Zardoub, 2019, Papadamou et al., 2020) and herding behaviour among them (Chen, 2020, Kyriazis, 2020a). More specifically a spectrum of studies has focused on volatility (Katsiampa, 2017, Chaim and Laurini, 2018, Corbet et al., 2018, Beneki et al., 2019), on market efficiency (Urquhart, 2016, Tiwari et al., 2018, Kyriazis, 2019a), on the impacts of investor attention (Subramaniam and Chakraborty, 2020) and the hedging properties of digital coins (Dyhrberg, 2016). A complete survey has also been conducted (Corbet et al., 2019) as well as bubble characteristics of cryptocurrency markets values (Kyriazis et al., 2020). Nevertheless, only a few studies have investigated the herding behaviour of cryptocurrencies (Bouri et al., 2019, Kallinterakis and Wang, 2019). This paper strengthens this stream of research. An important number of academic studies have investigated herding behaviour in cryptocurrency markets. Nevertheless, the majority of them employ conventional methodologies such as cross-sectional standard deviation (CSSD) and the cross-sectional absolute deviation (CSAD) estimations (Ballis and Drakos, 2019, da Gama Silva et al., 2019, Stavroyiannis and Babalos, 2019, Vidal-Tomás et al., 2019).

Herding behaviour in financial markets constitutes a frequent and highly-prominent phenomenon that is based on the irrational behaviour of economic agents. Herding phenomena are part of behavioural economics and finance that incorporate psychological factors such as framing, self-control, and justice into the analysis of economics (Thaler, 2016). It is supported that psychological factors are responsible for misperceptions and wrong decisions by the part of investors. Thereby, herding phenomena are responsible for inefficiency in financial markets and lead to the appearance of bull or bear tendencies as concerns the nominal values of financial assets (Fama, 1998, Shiller, 2003). It is supported by Shiller (2015) that herding behaviour in financial markets results in over-enthusiasm which leads to the appearance of bubbles in market prices.

According to Spyrou (2013), herding represents the non-rationality when economic agents mimic the irrational behaviour of other investors even when it is easily observable that the decisions these investors take are completely out of logic. Usual forms of herding phenomena constitute: (a) following the same direction as others when making investment decisions, (b) being based on historical decisions, (c) or fully imitating what other investors have done. Such behaviour is closely related to the higher intensity of the risk-seeking and speculative character of economic agents. Moreover, it is closely- and positively-linked with the ignorance and lack of management abilities by investors. Such phenomena become more detrimental during stressed market conditions and economic turmoil.

It should be noted that his paper adds to the existing literature about herding and concentration of market values by investigating price convergence and providing preliminary evidence of herding phenomena among digital currencies by the innovative methodologies of Phillips and Sul, 2007, Phillips and Sul, 2009 in the wide spectrum of 216 coins both in bull and bear markets. These models have been popular for measuring house price convergence (Churchill et al., 2018) or taxation (Regis et al., 2015). This methodology is highly innovative and in contrast to the majority of relevant studies about herding phenomena in cryptocurrency markets that employ the cross-sectional absolute deviation (CSAD) and the cross-sectional standard deviation (CSSD) methods. Convergence analysis in clusters differentiates from conventional herding analysis which is based on CSAD and CSSD measures. It should be emphasized that conventional methods reveal only whether the total cryptocurrencies examined follow a similar direction altogether. In contrast, price convergence in clubs enlightens as to whether leading cryptocurrencies form separate clusters of mimicking behaviour. Additionally, it is examined whether irrational behaviour prevails among the leaders of different clusters and it can be examined whether imitating other clusters’ behaviour enhances the concentration of cryptocurrencies into larger irrationally behaving groups. Thereby, not just an overall measure of mimicking is offered but a clearer view of the determinants (leaders) of such behaviour is provided.

To the best of our knowledge, this is the first study to employ such techniques to examine cryptocurrency price segmentation and herding behaviour in such a large spectrum of cryptocurrencies. This paper is related with Apergis et al. (2020) that focus on whether market microstructure drives convergence among major cryptocurrencies and support that it does. Moreover, they argue that the introduction of bitcoin futures contracts result in higher convergence for half of the cryptocurrencies examined. In contrast, our study centres interest on the herding phenomena among of a much wider range of digital currencies. Notably, the mimicking behaviour also of medium-cap and small-cap cryptocurrencies is under scrutiny.

Our findings reveal that herding is very intense during the bull market. Moreover, strong convergence behaviour is also estimated to exist during bear markets but in more clusters, as club convergence takes place among fewer cryptocurrencies that comprise each group Furthermore, clubs converge between them in pairs so secondary herding behaviour among clubs of digital currencies is detected. It is highly remarkable that this study about cryptocurrencies is the first one that does not compromise to just indicating whether there is overall herding behaviour or not. To be more precise, we examine whether convergence clubs are formed by the clustering method to provide a crystal-clear view of the specific clusters of transmitters and receivers of herding behaviour in the markets of digital currencies. Thereby, this is the first academic paper that enables not just to look into the cryptocurrency market in general but to make out which clusters of similar irrationality in behaviour are formed during extreme market conditions. To that end, the remainder of the paper is structured as follows. Section 2 presents the literature review on the herding behaviour of cryptocurrencies. Section 3 provides the data and methodology. Furthermore, Section 3 displays and analyses the results and provides the economic implications. Finally, Section 5 concludes and suggests avenues for further research.